Tutorials

Inference with pre-trained weights

Create a simple three layer network of integrate-and-fire neurons, densely connected with pre-trained weights and evaluate on MNIST test set

• Tutorial 1
  • Install
  • Download pre-trained weights
  • Install MNIST package
  • Build model

Training from scratch with e-prop

Directly train a simple three layer network of densely connected leaky integrate-and-fire neurons using the e-prop learning rule

• Tutorial 2
  • Install
  • Install MNIST package
  • Build model
  • Latency encoding
  • Network definition
  • Compilation
  • Training
  • Evaluate

Training from scratch with EventProp

Directly train a simple three layer network of densely connected leaky integrate-and-fire neurons using the EventProp learning rule

• Tutorial 3
  • Install
  • Install MNIST package
  • Build model
  • Latency encoding
  • Network definition
  • Compilation
  • Training
  • Evaluate

Training using TensorFlow and converting to an SNN using FewSpike

Create and train a simple convolutional neural network in Keras, convert to SNN using FewSpike and evaluate using mlGeNN

• Tutorial 4
  • Install
  • Train ANN
  • Evaluate ANN model
  • Build normalization dataset
  • Convert model
  • Compilation
  • Evaluate SNN models